CLPEX Chat Board

David Johnson wrote -

And to all you Poper fans while he may have some ideas on how science should be preformed he is not the be and end all of scientific thought or design

.

Thanks for bringing this up, David. Someone who understands science knows there is no 'be all and end all' aspect in science. Science is an open dialogue and not subject to absolutes, which is kind of what got us into trouble in the first place (Think 0% error rate) With the NAS report criticizing the scientific understanding of forensic practitioners and mentioning best practices, perhaps it's time to start a dialogue relative to their application in ACE-V. One topic that I have never seen discussed on this page is falsifiability in ACE-V.

As a scientific philosophy falsifiability carries a huge weight. As video killed the radio star, fasifiability killed historical quests for identification methodology(think Bertillionage). Furthermore from a philosophical standpoint, nothing can ever be proven. You don't know if something exists outside your observable domain that could render your hypothesis incorrect. This correlates to the semantic argument regarding individualization versus identification. "To the exclusion of all others" is logically and rationally indefensible.

How does this correlate to best practices in ACE-V? In my estimation, fasifiability comes into effect twice, once during C and the other during V. Procedurally, the comparison phase should begin to lead to a tentative hypothesis as agreement or disagreement begins to accumulate. As soon as you've reached a tentative hypothesis, it becomes time to test that hypothesis. Each further comparison on any level of detail should be predictive and should follow your tentative hypothesis. If the hypothesis is true, its predictive power should work both ways, from the latent to the known and from the known to the latent and you should see a continuation of agreement or disagreement depending on your tentative hypothesis.

It is at this point that falsification should come into effect as a normative methodology. You should attempt to falsify your tentative hypothesis. Doing so minimizes the incorporation of subjective agreement through circular reasoning and strengthens your tentative hypothesis. The limiting factor here is the number of trials you are able to perform this on. If you have very few comparative details, you might use them up just in coming to a tentative hypothesis and may not be able to incorporate falsifiability to arrive at sufficiency of identification or exclusion.

The second time falsification should come into effect is as an element of verification. Attempting to falsify the conclusion through a scrutiny of the observations and methods is going to end up giving the conclusion the strongest construct validity.

These are not topics I see covered in those who are considered overseers of our profession. The IAI has two positions on conclusions: (paraphrasing) We are not to testify to probabilities and we are not to testify to absolutes (So where does that leave us?). SWGFAST gives us a definition of Verification that it is merely confirming the conclusion and gives us a poor man's falsifiability tenet in the 'One Discrepency Rule'.

The bottom line is that confirmation strategies provide ambiguous information, where as fasifiability provides more thorough knowlege which is the ultimate goal of science in the first place.

I think this is where training and definitions fail. By making definitions simplistic, we are doing ourselves a disservice. If verification is simply confirming the conclusion, you enter into the infamous confirmation bias realm. For me, verification is a process by which I attempt to find information that DOES NOT confirm the original examiner's conclusion. When I am satisfied that I cannot, within the information present, find anything that does not confirm, then I can verify that it is an identification. No one told me that this was how I should approach a verification - I was taught that I should approach it as if I were doing the comparison for the first time. I think this is where confirmation bias appears...you already know it's supposed to be an ID - can you look at it critically, or do you give it a once-over and say it looks fine? Do I care what the crime is, what the detective wants, what the lawyers want? Nope. Do I take into account the training and experience of the person giving me the print to verify? Maybe. But, if I look at every verification as a cynic, and try to prove that it was not made by the person who made the prints on the card, then I feel like I'm approaching the task correctly.